about easyOCR fine tuning dataset #1431
Description
I’m working on fine-tuning a model for Persian digits, but I’m not sure whether the way I’ve prepared my dataset is correct.
Should all images in the dataset have a fixed height (for example, 32) and a variable width (e.g., between 165 and 190)?
Is it possible to find sample datasets that were used to train the arabic.pth model?
I’ve generated three types of images for my dataset, which I can share as examples.